The non-linear and lagged short-term relationship between rainfall and leptospirosis and the intermediate role of floods in the Philippines

Abstract

Background: Leptospirosis is a worldwide bacterial zoonosis. Outbreaks of leptospirosis after heavy rainfall and flooding have been reported. However, few studies have formally quantified the effect of weather factors on leptospirosis incidence. We estimated the association between rainfall and leptospirosis cases in an urban setting in Manila, the Philippines, and examined the potential intermediate role of floods in this association.

Methods/principal findings: Relationships between rainfall and the weekly number of hospital admissions due to leptospirosis from 2001 to 2012 were analyzed using a distributed lag non-linear model in a quasi-Poisson regression framework, controlling for seasonally varying factors other than rainfall. The role of floods on the rainfall-leptospirosis relationship was examined using an indicator. We reported relative risks (RRs) by rainfall category based on the flood warning system in the country. The risk of post-rainfall leptospirosis peaked at a lag of 2 weeks (using 0 cm/week rainfall as the reference) with RRs of 1.30 (95% confidence interval: 0.99-1.70), 1.53 (1.12-2.09), 2.45 (1.80-3.33), 4.61 (3.30-6.43), and 13.77 (9.10-20.82) for light, moderate, heavy, intense and torrential rainfall (at 2, 5, 16, 32 and 63 cm/week), respectively. After adjusting for floods, RRs (at a lag of 2 weeks) decreased at higher rainfall levels suggesting that flood is on the causal pathway between rainfall and leptospirosis.

Conclusions: Rainfall was strongly associated with increased hospital admission for leptospirosis at a lag of 2 weeks, and this association was explained in part by floods.

Fig 1. Assumed causal relationships among rainfall, flood and leptospirosis and other environmental factors.

Fig 2. Weekly number of admitted leptospirosis cases, cumulative rainfall, mean temperature and flood events.

(upper panel) Time series of the number of leptospirosis cases per week admitted to San Lazaro Hospital. (lower panel) Weekly cumulative rainfall (mm) (vertical bars), weekly mean temperature (°C) (solid line) and flood events (tick marks at the top).

Fig 3. Lagged relationships between rainfall and leptospirosis.

(A) Mean fitted relative risk surface over lag and weekly rainfall (flood-unadjusted model), (B) Cross-sectional plots of Fig 3(A) at constant lag / rainfall values with relative risk (RR) (solid red line) and 95% CIs (gray area) and (C) Cross-sectional plots at constant lag / rainfall values (flood-adjusted model) with relative risk (RR) (dotted blue line) and 95% CIs (gray area).

Fig 4. The relationship between flood occurrence and leptospirosis at lags 0 to 7 weeks.